Maybe someone here can help me out. The data from 1000 c.e. in the hockey stick graph presumeably didn’t come from digital thermometers and wasn’t recorded into some sort of database by dilligent grad students. It comes from tree ring data and earth cores or something?
OK, the modern data, say from 1800 or 1900 or something on forward does come from actual data sets recorded with old fashioned thermometers, or does it come from the same source as the 1000 c.e. data? Presumeably, the we can use the same types of measurements to determine global temperature today that were used to determine global temperatures from 1000 c.e., no? Then, how does the graph look if only data from the same types of measurements is used? Is there such a graph, and if so, where?
Further, how do we know the data gathered for 1000 c.e is accurate and comparable to today’s thermometer measurements, i.e. how do we know we’re comparing apples to apples here?
Further, how do we know the data gathered for 1000 c.e is accurate and comparable to today’s thermometer measurements, i.e. how do we know we’re comparing apples to apples here?
That’s what (a) multiple data sources and (b) confidence intervals are for, ben. Notice how wide the latter are the early data and how much narrower they are for the 20th century data? Wonderful thing, confidence intervals.
That didn’t really help much. See, I used data all from the same set, and I get this
Looks like two hockey sticks to me. Are there really two hockey sticks? Why or why not? Was that first hockey stick due to man made causes? Maybe it’s from the “little industrial revolution” of 500 c.e. or something.
Now then, if it’s fair to smooth out the old stuff to a greater degree than the new stuff becaues we have better measurements for the new stuff, that means that you can’t compare non-smoothed features in the new data to smoothed features in the old data. It’s not a fair comparison. The smoothing of the old stuff destroys the features that it might have had that we could have compared with the new stuff, no? If I’m wrong, why? Isn’t the hockey stick supposedly showing warming unprecedented in the last 2000 years? I don’t get it.
That didn’t really help much. See, I used data all from the same set, and I get this
Looks like two hockey sticks to me. Are there really two hockey sticks? Why or why not? Was that first hockey stick due to man made causes? Maybe it’s from the “little industrial revolution” of 500 c.e. or something.
Now then, if it’s fair to smooth out the old stuff to a greater degree than the new stuff becaues we have better measurements for the new stuff, that means that you can’t compare non-smoothed features in the new data to smoothed features in the old data. It’s not a fair comparison. The smoothing of the old stuff destroys the features that it might have had that we could have compared with the new stuff, no? If I’m wrong, why? Isn’t the hockey stick supposedly showing warming unprecedented in the last 2000 years? I don’t get it.
Ben, the point is that the 500CE points shouldn’t shift your estimate of the average anything like as much as the 2000CE points because the uncertainty surrounding them is much greater.
First, eyeballing and drawing in ‘best fit’ lines is a bad idea- it’s really easy to substitute what you think you know for what the data actually tell you. And this picture is a great example of that, there’s no way that the first ‘hockey stick’ is a good fit for that data…
[Tim wrote something about this during the election that touches on this topic : link ]
Second, you’re asking a lot of good questions, but they are ‘how do statistics and climatology work’ questions. That is, they aren’t going be answered in a sentence or two, you’re going to have to do some real studying to figure them out.
For example, do you understand what Kieran means by confidence intervals? The data aren’t being smoothed in the sense of rounding, so there isn’t information being “destroyed”… we’ve just got more accurate measurements in the very recent past, so the uncertainties are smaller.
To answer another part of your question, a great deal of the modeling that goes into these graphs involves correlating data sources with each other. For example, if we were using tree rings as a proxy measurement for rainfall, we could examine recent rainfall records and recent tree cores to make sure that we are indeed measuring apples and apples (ie if our tree ring-rainfall inference/model is incorrect, we might expect our calculated results from modern tree rings to differ from the very reliable meaures of modern rainfall).
Go check out www.realclimate.org; they’re dedicated to climate science, and so far the message board denizens have been very good about offering explainations.
I know that eyeballing a best fit is dumb. I did it to make a point, which you acknowledged in your reply. I know the first hockey stick is no good, but for that set of data, neither is the second. The second hockey stick is supported by modern data, and that’s all well and good, but you can’t say that there wasn’t a hockey stick in the past, because you don’t know. So the hockey stick graph doesn’t show anything historically unique, except maybe increased accuracy in temperature readings with time.
Please someone tell me you know what I’m trying to say. Don’t refute points that I’m not making.
First, it wasn’t at all clear that you were drawing a fit for rhetorical purposes only…
Second, I don’t know where you got the data you’re graphing- are you graphing a single proxy (eg tree rings, ice cores, corals, lake and ocean sediments, tree pollen)? Some composite of several proxies? And, of course, whatever proxy/ies was used would’ve had to have been calibrated (the raw data aren’t very useful), and calibrations differ…
So, part of the answer depends on that. The rest basically leads back to the uncertainty issue already pointed out by Kieran- ie yes, it’s possible that there was a previous ‘hockey stick’-like increase in temperatures in the last thousand years, but it’s very, very, very unlikely. We don’t need to extend modern methods into the past to see that, all we need to do is determine the uncertainty of the measurements we’ve already got. But we’ll never know in some absolute sense; science does not allow for that level of certainty.
Check out link for a good discussion of the hockey stick.
The reason that it’s very, very, very unlikely has to do with statistics. Let’s say that you took a group of 100 people, and selected 10 at random. You measure the heights of those 10 people, and figure that the average height of the group is 5′8″.
But how sure are you that this is the correct number? Better, how sure are you that the actual average of the entire group is between, say, 5′7″ and 5′9″? Given certain assumptions (eg random selection, regular distribution (ie the class isn’t made up of 95 pygmies and 5 NBA centers)), one can calculate the probability that the average does indeed fall within the range. Of course, there is a small chance that you’ve picked the 10 shortest or tallest people in the room- ergo, the associated probability.
So, it’s possible in that scenario that the average height is actually 6′2″, but it is very unlikely.
If you’ll follow the link I provided earlier, you’ll see several diagrams of the Mann and Jones 2003 (NH) data with the associated uncertainties. Bearing in mind that the actual line could lie outside of the uncertainty range (ergo, there could be a hockey stick back there someplace)- but it probably doesn’t.
Also see link for their “myths about the hockey stick”. If we only had one data set that supported the idea, the scientific conclusion would be much more tentative. But it’s been supported by every peer-reviewed subsequent study, and a number of other analyses- so scientists are much more confident about the conclusion.
Wow, this IS more accurate than the hockey stick ! More honest too. And, maybe this time, doesn’t rely almost entirely on a set of bristlecone pine growth records that ARE NOT RELATED TO TEMPERATURE. Do you actually understand the effects of the mathematical contortions Mann went through to produce the “stick” ? No ? I thought not, try working through it first until you do understand, then comment, please.
January 28th, 2005 at 5:32 am
It may be a parody, but I think you still made fewer errors and more sense than the average TCS column.
January 28th, 2005 at 7:12 am
Maybe someone here can help me out. The data from 1000 c.e. in the hockey stick graph presumeably didn’t come from digital thermometers and wasn’t recorded into some sort of database by dilligent grad students. It comes from tree ring data and earth cores or something?
OK, the modern data, say from 1800 or 1900 or something on forward does come from actual data sets recorded with old fashioned thermometers, or does it come from the same source as the 1000 c.e. data? Presumeably, the we can use the same types of measurements to determine global temperature today that were used to determine global temperatures from 1000 c.e., no? Then, how does the graph look if only data from the same types of measurements is used? Is there such a graph, and if so, where?
Further, how do we know the data gathered for 1000 c.e is accurate and comparable to today’s thermometer measurements, i.e. how do we know we’re comparing apples to apples here?
January 28th, 2005 at 7:15 am
You mean, that “big dipper” graph actually appeared in a TCS article? Please tell me it didn’t.
January 28th, 2005 at 2:56 pm
Further, how do we know the data gathered for 1000 c.e is accurate and comparable to today’s thermometer measurements, i.e. how do we know we’re comparing apples to apples here?
That’s what (a) multiple data sources and (b) confidence intervals are for, ben. Notice how wide the latter are the early data and how much narrower they are for the 20th century data? Wonderful thing, confidence intervals.
January 28th, 2005 at 4:49 pm
That didn’t really help much. See, I used data all from the same set, and I get this
Looks like two hockey sticks to me. Are there really two hockey sticks? Why or why not? Was that first hockey stick due to man made causes? Maybe it’s from the “little industrial revolution” of 500 c.e. or something.
Now then, if it’s fair to smooth out the old stuff to a greater degree than the new stuff becaues we have better measurements for the new stuff, that means that you can’t compare non-smoothed features in the new data to smoothed features in the old data. It’s not a fair comparison. The smoothing of the old stuff destroys the features that it might have had that we could have compared with the new stuff, no? If I’m wrong, why? Isn’t the hockey stick supposedly showing warming unprecedented in the last 2000 years? I don’t get it.
January 28th, 2005 at 4:53 pm
That didn’t really help much. See, I used data all from the same set, and I get this
Looks like two hockey sticks to me. Are there really two hockey sticks? Why or why not? Was that first hockey stick due to man made causes? Maybe it’s from the “little industrial revolution” of 500 c.e. or something.
Now then, if it’s fair to smooth out the old stuff to a greater degree than the new stuff becaues we have better measurements for the new stuff, that means that you can’t compare non-smoothed features in the new data to smoothed features in the old data. It’s not a fair comparison. The smoothing of the old stuff destroys the features that it might have had that we could have compared with the new stuff, no? If I’m wrong, why? Isn’t the hockey stick supposedly showing warming unprecedented in the last 2000 years? I don’t get it.
January 28th, 2005 at 4:56 pm
hmmm, I guess pics aren’t allowed. It showed up when I previewed the post.
January 28th, 2005 at 7:31 pm
Ben, the point is that the 500CE points shouldn’t shift your estimate of the average anything like as much as the 2000CE points because the uncertainty surrounding them is much greater.
January 28th, 2005 at 7:44 pm
Ben,
First, eyeballing and drawing in ‘best fit’ lines is a bad idea- it’s really easy to substitute what you think you know for what the data actually tell you. And this picture is a great example of that, there’s no way that the first ‘hockey stick’ is a good fit for that data…
[Tim wrote something about this during the election that touches on this topic : link ]
Second, you’re asking a lot of good questions, but they are ‘how do statistics and climatology work’ questions. That is, they aren’t going be answered in a sentence or two, you’re going to have to do some real studying to figure them out.
For example, do you understand what Kieran means by confidence intervals? The data aren’t being smoothed in the sense of rounding, so there isn’t information being “destroyed”… we’ve just got more accurate measurements in the very recent past, so the uncertainties are smaller.
To answer another part of your question, a great deal of the modeling that goes into these graphs involves correlating data sources with each other. For example, if we were using tree rings as a proxy measurement for rainfall, we could examine recent rainfall records and recent tree cores to make sure that we are indeed measuring apples and apples (ie if our tree ring-rainfall inference/model is incorrect, we might expect our calculated results from modern tree rings to differ from the very reliable meaures of modern rainfall).
Go check out www.realclimate.org; they’re dedicated to climate science, and so far the message board denizens have been very good about offering explainations.
January 28th, 2005 at 9:50 pm
Steve, David (”Greening Earth Society”) Wojick actually did present a graph exactly like Tim’s “big dipper” - here is one of its sources:
link
Tim, great piece! I love it!
Jeff
January 29th, 2005 at 4:33 am
Carleton,
I know that eyeballing a best fit is dumb. I did it to make a point, which you acknowledged in your reply. I know the first hockey stick is no good, but for that set of data, neither is the second. The second hockey stick is supported by modern data, and that’s all well and good, but you can’t say that there wasn’t a hockey stick in the past, because you don’t know. So the hockey stick graph doesn’t show anything historically unique, except maybe increased accuracy in temperature readings with time.
Please someone tell me you know what I’m trying to say. Don’t refute points that I’m not making.
January 30th, 2005 at 3:25 am
Ben,
First, it wasn’t at all clear that you were drawing a fit for rhetorical purposes only…
Second, I don’t know where you got the data you’re graphing- are you graphing a single proxy (eg tree rings, ice cores, corals, lake and ocean sediments, tree pollen)? Some composite of several proxies? And, of course, whatever proxy/ies was used would’ve had to have been calibrated (the raw data aren’t very useful), and calibrations differ…
So, part of the answer depends on that. The rest basically leads back to the uncertainty issue already pointed out by Kieran- ie yes, it’s possible that there was a previous ‘hockey stick’-like increase in temperatures in the last thousand years, but it’s very, very, very unlikely. We don’t need to extend modern methods into the past to see that, all we need to do is determine the uncertainty of the measurements we’ve already got. But we’ll never know in some absolute sense; science does not allow for that level of certainty.
Check out link for a good discussion of the hockey stick.
January 30th, 2005 at 7:45 am
why?
January 30th, 2005 at 8:19 am
The data set is one that Michael Mann used. The source is from his website here. I used only the summed north and south hemisphere data.
January 31st, 2005 at 3:35 am
The reason that it’s very, very, very unlikely has to do with statistics. Let’s say that you took a group of 100 people, and selected 10 at random. You measure the heights of those 10 people, and figure that the average height of the group is 5′8″.
But how sure are you that this is the correct number? Better, how sure are you that the actual average of the entire group is between, say, 5′7″ and 5′9″? Given certain assumptions (eg random selection, regular distribution (ie the class isn’t made up of 95 pygmies and 5 NBA centers)), one can calculate the probability that the average does indeed fall within the range. Of course, there is a small chance that you’ve picked the 10 shortest or tallest people in the room- ergo, the associated probability.
So, it’s possible in that scenario that the average height is actually 6′2″, but it is very unlikely.
If you’ll follow the link I provided earlier, you’ll see several diagrams of the Mann and Jones 2003 (NH) data with the associated uncertainties. Bearing in mind that the actual line could lie outside of the uncertainty range (ergo, there could be a hockey stick back there someplace)- but it probably doesn’t.
Also see link for their “myths about the hockey stick”. If we only had one data set that supported the idea, the scientific conclusion would be much more tentative. But it’s been supported by every peer-reviewed subsequent study, and a number of other analyses- so scientists are much more confident about the conclusion.
February 4th, 2005 at 7:11 am
Wow, this IS more accurate than the hockey stick ! More honest too. And, maybe this time, doesn’t rely almost entirely on a set of bristlecone pine growth records that ARE NOT RELATED TO TEMPERATURE. Do you actually understand the effects of the mathematical contortions Mann went through to produce the “stick” ? No ? I thought not, try working through it first until you do understand, then comment, please.
February 4th, 2005 at 5:58 pm
Do you actually understand the effects of the mathematical contortions Mann went through to produce the “stick” ?
WOW, do you actually understand the term “proxy”? No? I thought not.
This is a dead parrot. It has gone to meet its maker.
D